JPH0451035B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a voice detection method and circuit for determining the presence or absence of a voice signal.

(Prior art and its problems) Voice detection circuits are mainly DSI (Digital Speech
It is an abbreviation for Interpolation. ) incorporated into the device
Used to determine whether an audio signal is present on the input channel to the DSI device.

Regarding DSI equipment, for example, there is a document published in March 1976, COMSAT TECHNICAL REVIEW magazine vol.6 No.
Please refer to the article ``Digital Speech Interpolation'' by S.J. Campanella, published on pages 127 to 158 of 1, Vol.

Conventionally, the level detection method is known as a method with simple hardware scale and clear detection logic, but this method detects the signal energy (power and amplitude) of the input signal and then compares it with a threshold. This is to determine the presence or absence of a signal. Furthermore, among the sound detectors using the level detection method, the fixed threshold sound detector, which compares the amplitude of the input signal with a predetermined threshold, has the simplest hardware scale and is the most reliable sound detector. Are known.

Next, the principle of this fixed threshold type voice detector will be explained with reference to the drawings.

FIG. 1 is a block diagram showing the principle of a fixed-threshold sound detector, in which there is a signal input terminal 1, an amplitude threshold input terminal 2, an amplitude comparison circuit 3, an accumulation circuit 4, and +1 and -1 input signals to the accumulation circuit. increase/decrease control lines and 6,5 voice detection flipflop 7, voice detection flipflop set/reset control lines 8 and 9, voice detection result output terminal 10.
It is made up of In this case, the accumulation circuit can be replaced with a reversible counter (up-down counter).

In FIG. 1, the input signal input from terminal 1 is input to the amplitude comparator circuit 3 for each sampling period.
It is compared with a predetermined amplitude threshold (THa) input from terminal 2. As a result, if the input signal amplitude is greater than the amplitude threshold, the contents of the accumulator circuit 4 are incremented by one using the accumulator increment control line 5. Conversely, when the input signal amplitude is smaller than the amplitude threshold, the contents of the accumulator circuit 4 are decreased by 1 using the accumulator decrease control line 6. However, the contents of the accumulator circuit are designed not to take a negative value.

As audio signals arrive and more inputs exceed the amplitude threshold, the contents of the accumulator circuit increase sequentially. Of course, if an input below the amplitude threshold is added during that time,
The contents of the reversible counter are decremented by one. In this way, when the contents of the accumulator circuit reach the preset duration threshold (THt), the voice detection flip-flop 7 is set using the voice detection flip-flop set control line 8, and the voice is detected. The result is output from the terminal 10.

Also, when no voice is detected, which is indicated by the content of the accumulator circuit 4 becoming 0, for example, the voice detection flip-flop 7 is reset using the voice detection flip-flop reset control line 9; The result is output from terminal 10, but is generally reset after a certain period of time. This is called a hangover, and is created because the ear is sensitive to disconnections between words or phrases during a call, and the length of the hangover is 100 to 100 minutes.
It is about 250ms.

For further understanding, an explanation will be given by taking as an example a case where a signal indicated by 11 in a of FIG. 2 is input to the fixed threshold type voice detector shown in FIG. 1.

In FIG. 2, a is the input signal 11 and the amplitude threshold 1
2. Figure b shows the contents 13 of the accumulation circuit and the duration threshold 14, and figure c shows the voice detection result output 15.

First, when the input signal 11 is input from the terminal 1, it is compared with the amplitude threshold value 12 by the amplitude comparison circuit 3 every sampling period Ts. As can be seen from Figure 2, the amplitude of the input signal becomes larger than the amplitude threshold only at time t _a1 , so the contents of the accumulation circuit 1
3 starts to become 1 at time t _a1 (Fig. 2b), and thereafter increases by 1 until time t _a2 . As a result, at time _ta1 , the content 13 of the accumulator circuit becomes larger than the duration threshold 14, which means that voice has been detected, and the output 15 becomes 1. By the way, at time t _a3 , the amplitude of the input signal 11 becomes smaller than the amplitude threshold 12, so the content 13 of the accumulation circuit decreases by 1, and at time t _a2 ,
Since the duration becomes smaller than the threshold 14, it is determined that the audio signal has disappeared, and a hangover is added for the above-mentioned reason, and after the hangover ends, the output is 1.
5 becomes 0. T _H in c of FIG. 2 indicates the hangover time.

Although the fixed-threshold sound detector described above has a simple hardware scale, it has the drawback that once the threshold is set, any noise that is above the threshold will be detected.

A variable threshold type voice detector is known as a device that improves the drawbacks of such a fixed threshold type voice detector. DESCRIPTION OF THE PREFERRED EMBODIMENTS A conventional variable threshold sound detector will be described in detail below with reference to the drawings. FIG. 3 shows a conventional variable threshold audio detector, which includes an input terminal 20, an even bit inversion circuit 21, a code conversion circuit 22, a rectification circuit 23, a power calculation circuit 24, and a first threshold generation circuit 2.
5, a level detection circuit 26, an accumulation circuit 27, a comparison circuit 28, a second threshold generation circuit 29, a reversible counter 30, a counter setting circuit 37, a determination circuit 32, and an output terminal 33.

For example, the International Telegraph and Telephone Consultative Committee, (CCITT;
Comit'e Consultatif International Te'leg−
8-bit A-Law code (Orange Book Vol.-2,
See pp409-410. ) is input from the input terminal 20 as an example. Normally, the even numbered bits of an A-Law code signal transmitted over a telephone line are inverted when viewed from the MSB (abbreviation for Most Significant Bit) side, so the even numbered bits of the input signal are inverted by the even numbered bit inversion circuit 21. It is inverted and returned to the signal before being transmitted. The A-Law code signal returned to the original state is sent to the code conversion circuit 22, and as shown in FIG. All bits are inverted and converted into an 8-bit two's complement sign signal, which is input to the rectifier circuit 23. The rectifier circuit 23 converts this input signal into an absolute value signal (signal representing only the magnitude), one of which is sent to the power calculation circuit 24 and the other to the level detection circuit 26.

The power calculation circuit 24 extracts noise contained in the input signal and calculates the average power of the noise. Specifically, when no voice is detected (for example, when the output of the comparison circuit 28 described later is 0), all input signals are regarded as noise, and when voice is detected (for example, when the output of the comparison circuit 28 described later is 0), all input signals are regarded as noise. Even if the output is 1), signals below a predetermined level are considered to be noise, and the average power of the noise is calculated by inputting this noise to a low-pass filter, and the result is used as the first filter. The threshold value generation circuit 25 of FIG. Therefore, the audio signals that are excluded when calculating the average power of noise are those whose output from the comparison circuit described below is 1.
and has a signal level equal to or higher than a predetermined level. First threshold generation circuit 2
5, by multiplying the output from the power calculation circuit 24 by a constant, the first threshold value (TH1) used in the level detection circuit 26 is set to 6 dB from the first threshold value.
A second threshold value (TH2) is set at a high point and sent to the level detection circuit 26.

The level detection circuit 26 uses the output of the rectification circuit 23 and the first threshold value sent from the first threshold generation circuit 25.
If the output of the rectifier circuit is larger than the second threshold, the probability that the input signal is an audio signal is +2, and the difference between the first threshold and the second threshold is +2. If it is between, the probability that the input signal is a voice signal and the probability that it is noise are approximately equal, or the former is slightly higher, so +
1. If it is smaller than the first threshold, there is a high probability that the input signal is noise, so -1 is output. The accumulation circuit 27 accumulates the output of the level detection circuit 26 and sends the accumulated value to the comparison circuit 28. In the comparison circuit 28, a second threshold generation circuit 2, which will be described later,
The third threshold value (TH3) output from 9 is compared with the cumulative value, and if the latter is larger than the former, it is determined that the input signal is an audio signal, and +1 is
Furthermore, if the former is larger than the latter, it is determined that the input signal is noise, and 0 is output. In the second threshold generation circuit 29, two thresholds with different levels are prepared as the third threshold (TH3) for audio signal judgment used in the comparison circuit 28, and the output of the comparison circuit 28 is When the output of the comparator circuit 28 is 0, a high level third threshold (TH3H) is generated, and when the output of the comparator circuit 28 is 1, a low level third threshold (TH3L) is generated.
Send to.

The reversible counter 30 inputs the output of the comparison circuit 28, and when the input signal is 1, the contents of the counter are increased by 1, and when the input signal is 0, the contents of the counter are decreased by 1, and the output of the comparison circuit is accumulated. ing. Further, the counter setting circuit 31 monitors the output of the comparison circuit, detects the point in time when the output changes from 1 to 0, and sets the contents of the reversible counter 30 to a predetermined value at that point. . In the determination circuit 32, the contents of the reversible counter 30 are set to a predetermined value (usually 0 is used).
1 if an audio signal is detected.
is output to the outside via the output terminal 33. Of course, if it is small, 0 is output, but doing so will also result in the above-mentioned hangover.

The circuit shown in FIG. 5 can be used as the power calculation circuit 24 and the second threshold generation circuit 25 in FIG. unit 53, OR circuit 54, multipliers 55, 56,
57, 58, multiplicand input terminals 59, 60, 61,
It consists of a multiplicand selector 62, 63, an adder 65, limiters 66, 67, a memory 68, a first threshold output terminal 69, and a second threshold output terminal 70. The absolute value input signal is input from an input terminal 50, one of which is sent to a multiplier 55 and the other to a comparator 53. In the comparator 53, the input signal is compared with the noise judgment level input from the input terminal 51, and if the former is larger than the latter, 0,
If it is smaller, +1 is output, and the OR circuit 54 performs the OR of the output signal of the comparator 53 and the inverted signal of the output signal from the comparison circuit 28.
+1 is output when at least one of them is +1, and serves as a control signal for the multiplier 56 and a selection control signal for the multiplicand selector 64. In the multiplicand selector 64, when the selection control signal is +1, the multiplicand input from the multiplicand input terminal 59 is selected, and when it is 0, the multiplicand input from the multiplicand input terminal 60 (currently 0 is used. ) is selected and becomes the multiplicand of the multiplier 55.

Further, in the multiplier 55, the product of the absolute value input signal and the multiplicand selected as described above is calculated and sent to the adder 65. On the other hand, in the multiplier 56, the multiplicand memory 6 input from the multiplicand input terminal 61
The product is multiplied with the contents of 8 and sent to adder 65.
However, when the output of the OR circuit 54 is 0, this multiplication is not performed and the contents of the memory 68 are output as they are. Then, an adder 65 adds the output of the multiplier 55 and the output of the multiplier 56, and the result is stored in a memory 68 via a limiter 66. At the same time, the output of the limiter 66 is multiplied by the multiplicand input from the multiplicand input terminal 62 by the multiplier 57, and is outputted from the output terminal 69 via the limiter 67 as a first threshold value (TH1). Ru.

Further, the output of the limiter 67 is multiplied by the multiplicand (currently 2 is used) inputted from the multiplicand input terminal 63 in a multiplier 58, and the product is determined as a second threshold value (TH2) at the output terminal 70. It is output from

Here, the limiters 66 and 67 are used to quickly adjust the threshold value by limiting the contents of the memory 68 and the variable range of the threshold value (TH1), and to limit the reception sensitivity and emotional level range of the audio detector. This is to ensure immunity from noise.

Note that the power calculation circuit 24 calculates the noise level by passing the absolute value signal through a first-order low-pass filter as described above, but this is because the amplitude distribution is
The power P obtained by taking the absolute value of noise with a Gaussian distribution and a variance of σ ² and passing it through a first-order low-pass filter is approximately equal to the standard deviation σ as shown in the following equation.
This is because the value is proportional to .

Here, if we set x ² = y, then 2xdx = dy, so ∫ ^∞ _-∞ xexp (-x ² / 2σ ² ) dx = ∫ ⁰⁰ ₀ 1/2xep (-y / 2σ ² ) dy = σ ² , so equation (1) becomes as follows.

Therefore, P≈0.8σ (3) Therefore, it can be seen that by performing the above processing, a value approximately proportional to the standard deviation σ of the noise can be obtained at the output of the first-order low-pass filter.

Further, as shown in FIG. 3, a second threshold generation circuit 29 is provided, two third thresholds (TH3) are prepared to be used in the comparison circuit 28, and the output of the comparison circuit 28 is used as a selection signal. When is 1, a low level third threshold (TH3L) is selected and when it is 0, a high level third threshold (TH3H) is selected and used. This is because hysteresis is provided in the output of the comparison circuit 28. Excessive ON of the audio detector due to
This is to avoid turning off.

The reversible counter 30, counter setting circuit 31, and determination circuit 32 used in FIG.
The circuit shown in the figure can be used, including an input terminal 71, a one-sample delay circuit 72, an AND circuit 73, a counter setting value input terminal 74, a reversible counter 75, and a comparison circuit 7.
6, a threshold input terminal 77 and an output terminal 78, and 30, 31, and 32 surrounded by broken lines indicate a reversible counter, a counter setting circuit, and a determination circuit shown in FIG. 3, respectively. One side of the input signal input from the input terminal 71 is sent to the reversible counter 75.
and the other one is sent to a one-sample delay circuit 72 and an AND circuit 73.

The AND circuit 73 multiplies the signal obtained by inverting the current input signal and the input signal one sample time ago.
The result is sent to the reversible counter 75. The reversible counter 75 increases the contents of the counter by 1 when the input signal is 1, and decreases the contents of the counter by 1 when the input signal is 0. 1
When the value changes from 0 to 0, the contents of the counter are forcibly set to a predetermined value input from the counter setting value input terminal 74. The comparator 76 compares the threshold value input from the threshold input terminal 77 (actually, 0 is used) with the content of the counter output from the reversible counter 75, and outputs 1 if the content of the counter is larger. It is output to the outside via the terminal 78.

As mentioned above, in the conventional variable threshold voice detection circuit,
Since signal processing is performed using PCM codes, the hardware size does not increase, a threshold value corresponding to the noise level can be obtained, and there is strong immunity against spurious signals, and the maximum and minimum values of the threshold value Although there is an advantage in that the reception sensitivity and emotional level range can be set arbitrarily by specifying the When a plosive consonant such as "ch" continues, the speech detector detects only the "su ^- " part and treats "ch" as noise and does not detect it (word-ending truncation), or even if it detects it, it does not detect the "ch" part. Most of the words are lost (dropped out). Also, when pronouncing the English word "end" with a slight distance between "en" and "d", the pronunciation level of "en ^- " and "d" is stronger than " ^d ". ”, and even in such cases, speech detectors were often unable to detect the ending of the word.

(Object of the Invention) An object of the present invention is to provide a speech detector that is less likely to cause dropouts in the middle of words or truncations at the end of words as described above.

(Structure of the Invention) According to the present invention, the magnitude determination result between the input signal inputted at each sample time and the first threshold value and the second threshold value that vary depending on the noise level included in the input signal is determined numerically. The signal power of the audio signal included in the input signal is determined by accumulating the converted values, and the presence or absence of the audio signal is determined by comparing the signal power with a third threshold value, and the determination result changes from the presence of audio to the absence of audio. In the voice detection method for obtaining an output by adding the determination result as a voiced interval at a predetermined time when the voice detection output exceeds the threshold value, two values with different levels are prepared as the third threshold, and the voice detection output is When the voice detection output indicates the presence of a voice signal, a low level value is used, and when the voice detection output indicates that there is no voice signal, a high level value is used to perform voice detection, and a low level threshold is used even in the added sound section. A characteristic voice detection method is obtained.

Further, according to the present invention, there is provided a means for generating a first threshold value and a second threshold value according to noise power included in an input signal input at each sample time; a comparison means for comparing the output of the accumulating means with a third threshold and outputting a signal indicative of the comparison result; Detects the time when the state changes from "large" to "small" and indicates "sound present" only for a predetermined period of length immediately after this change time and during the period when the output of the comparison means is "large" A determination means for outputting a determination signal, and a predetermined high value and a predetermined low value are input, and when the determination signal supplied from the determination means indicates "audio presence", the low value is inputted,
Otherwise, there is obtained a voice detection circuit characterized in that it includes at least means for supplying the high value as the third threshold value to the comparison means.

(Embodiment) The present invention has the above-described configuration, and uses the second threshold at the same low level as the voiced section even when a hangover is added, thereby reducing dropouts and endings of words. There is.

The present invention will be explained in detail with reference to the drawings.
FIG. 7 shows an embodiment of the present invention, in which the input terminal 2
0, even bit inversion circuit 21, code conversion circuit 2
2. Rectifier circuit 23, power calculation circuit 24, first threshold generation circuit 25, level detection circuit 26, accumulation circuit 27, comparison circuit 28, second threshold generation circuit 29,
It is composed of a reversible counter 30, a counter setting circuit 31, and an output terminal 33. Here, an even bit inversion circuit 21, a code conversion circuit 22, a rectification circuit 23, a power calculation circuit 24, a first threshold generation circuit 25, a level detection circuit 26, an accumulation circuit 27, a comparison circuit 28, a second threshold generation circuit 29, reversible counter 30, and counter setting circuit 32 are the same as those shown in FIG. The differences between the conventional sound detector shown in FIG. 3 and the sound detector of the present invention shown in FIG.
Whereas the signal input to the control terminal c of the second threshold generation circuit 29 in FIG. The point is that the signal input to terminal c has changed to the signal at output terminal 33. Hereinafter, the details of the sound detection mechanism from the terminal 20 to the cumulative calculation circuit 27 will be given to the explanation of the operation of the conventional sound detector described in FIG. threshold generation circuit 29, reversible counter 30, counter setting circuit 3
1, an example of its operation was shown. Figure 8 a-d
This will be explained in detail with reference to . Now, the audio detector receives the signals shown in FIG. 8a (audio signals 130 and 13).
Waveform 13 considers the case where 1 (including 1) is input.
0 arrives, and in the level detection circuit 26, the first
and the second threshold, and an output like the waveform 132 shown as b is obtained as the output of the accumulation circuit 27. In the comparator circuit 28, the output waveform 132 of the accumulation circuit 27 and the third threshold are obtained. The change waveform 134 is compared.

Until time T ₁ , the third threshold change waveform 134 is larger than the output waveform 132 of the accumulation circuit 27, so no audio signal is detected; however, at time T ₁ , the latter becomes larger, so the output of the comparison circuit 28 Waveform 1
35 becomes 1, and the content waveform 137 of the reversible counter 30 for holding the output also begins to increase.Therefore, the output waveform 138 of the output terminal 33 also becomes 1, indicating that an audio signal has been detected.

Therefore, when the output waveform 138 of the output terminal 33 becomes 1, the second threshold generation circuit 29 selects the third threshold at a low level, so that it is set to a low value as shown in FIG. 8b.

Thereafter, the amplitude of the input waveform 130 shown in _FIG .
becomes 0 as shown in FIG. 8c.

However, the content waveform 1 of the reversible counter 30
37 is a data setting circuit 31 as shown in FIG. 8c.
Since the data for hangover is set, hangover is added.

When hangover is added, terminal 2
When the waveform 131 shown in FIG. 8a arrives at 0, the time
At T ₃ , the output waveform 133 of the accumulation circuit 27 changes to the waveform 1 shown in FIG. 8b output from the second threshold generation circuit 29.
34, the output of the comparator circuit 28 has a waveform of 1 as shown in FIG.
37 increases again. Then, at time T ₄ when the output of the accumulator circuit 27 becomes smaller than the waveform 133 shown in FIG.
Since the waveform 136 in FIG. c becomes 0, a hangover is added again here.

When the output terminal 33 returns to 0 at time _T5 , the second
The threshold value generating circuit 29 generates the changing waveform 134 in FIG. 8b.
It is set to a high value again at time _T5 , as shown in .

Therefore, if the second threshold generation circuit 29 was controlled by the comparison circuit 28 as before, the waveform 131 shown in _FIG . Waveform 131 in Figure 8a
It is also possible to detect Therefore, it can be seen that the present invention improves the ability to prevent omissions of words and phrases and omission of endings.

(Effects of the Invention) As explained in detail above, according to the present invention, the ability to prevent words and phrases from being omitted and erroneous endings from being omitted can be improved, especially when the present invention is used in a DSI device. The effect is extremely large.

[Brief explanation of the drawing]

Figure 1 is a block diagram showing a conventional voice detector.
Figures 2a, b, and c are diagrams showing the waveforms of each part in Figure 1, Figure 3 is a block diagram showing another conventional voice detector, Figure 4 is a diagram showing the code conversion method, and Figure 5. ,
FIG. 6 is a diagram showing a part of FIG. 3; FIG. 7 is a block diagram showing a variable threshold type speech detector according to the present invention;
8A to 8D are diagrams for explaining the operation of FIG. 7. In the figure, 20 is an input terminal, 21 is an even bit inversion circuit, 22 is a code conversion circuit, 23 is a rectifier circuit, 26 is a level detection circuit, 27 is an accumulation circuit, 2
8 is a comparison circuit, 29 is a second threshold generation circuit, 30
is a reversible counter, 31 is a counter setting circuit, 32
is a determination circuit, and 33 is an output terminal.

Claims (1)

[Claims] 1. A value obtained by quantifying the magnitude determination result between an input signal input at each sample time and a first threshold value and a second threshold value that vary depending on the noise level included in the input signal. The signal power of the input signal is cumulatively obtained, the signal power is compared with a third threshold value, the presence or absence of an audio signal is determined, and if the determination result changes from the presence of an audio signal to the absence of an audio signal, a predetermined signal power is determined. In a voice detection method that obtains a voice detection output by adding time as a sound interval to the determination result, two values with different levels are prepared as the third threshold, and when the voice detection output indicates the presence of a voice signal. Voice detection using a low level value, and when the voice detection output indicates that there is no voice signal, voice detection is performed using a high level value, and a low level threshold is used even in the added sound section. Method. 2. Means for generating a first threshold value and a second threshold value according to noise power included in an input signal input at each sampling time, and comparing the input signal with the first threshold value and the second threshold value. means for digitizing and accumulating the comparison results; comparing means for comparing the output of the accumulating means with a third threshold value and outputting a signal indicating the presence or absence of an audio signal; A determination that detects the time when the state changes to "absent" and outputs a determination signal indicating "audio presence" only during a predetermined length period immediately after the time of change and during the period during which the output of the comparison means is the audio signal present. means and
A predetermined high value and a low value are input, and when the judgment signal supplied from the judgment means indicates "sound present", the low value is used; otherwise, the high value is used as the first value. 3. A voice detection circuit comprising at least means for supplying a threshold value of 3 to the comparison means.